(2) Health Economics and Decision Science (HEDS) School Of Health And. Discussion Paper. Related. Research.. The Sheffield Type 1 Diabetes Policy Model Praveen Thokala, Jen Kruger, Alan Brennan, Hasan Basarir, Alejandra of DP. of DP Pandor, Mike Duenas, Abdullah. Gillett, Jackie Elliot, Simon Heller. DP 13/05. This series is intended to promote discussion and to provide information about work in progress. The views expressed are those of the authors, and therefore should not be quoted without their permission. However, comments are welcome and we ask that they be sent direct to the corresponding author..
(3) HEDS Discussion Paper No.13.05 The Sheffield Type 1 Diabetes Policy Model. Praveen Thokala1, Jen Kruger1, Alan Brennan1, Hasan Basarir1, Alejandra Duenas1, Abdullah Pandor1, Mike Gillett1, Jackie Elliot2, Simon Heller2 1. Health Economics and Decision Science, School of Health and Related Research, University of Sheffield. 2. Academic Unit of Diabetes, Endocrinology & Metabolism, Department of Human Metabolism, The University of Sheffield, The Medical School, Beech Hill Road, Sheffield S10 2RX, UK. Disclaimer: This series is intended to promote discussion and to provide information about work in progress. The views expressed in this series are those of the authors, and should not be quoted without their permission. Comments are welcome, and should be sent to the corresponding author.. This paper is also hosted on the White Rose Repository: http://eprints.whiterose.ac.uk/. White Rose Research Online firstname.lastname@example.org. -1-.
(4) The Sheffield Type 1 Diabetes Policy Model Praveen Thokala1, Jen Kruger1, Alan Brennan1, Hasan Basarir1, Alejandra Duenas1, Abdullah Pandor1, Mike Gillett1, Jackie Elliot2, Simon Heller2 1. Health Economics and Decision Science, School of Health and Related Research, University of Sheffield. 2. Academic Unit of Diabetes, Endocrinology & Metabolism, Department of Human Metabolism, The University of Sheffield, The Medical School, Beech Hill Road, Sheffield S10 2RX, UK. Correspondence to: Jen Kruger Health Economics and Decision Science, University of Sheffield, Regents Court, 30 Regent Street, Sheffield S1 4DA, UK +44 (0)114 222 5207 email@example.com. Key words: type 1 diabetes; health economic model; simulation; cost-effectiveness. Funding information: This project was funded by the UK National Institute for Health Research (NIHR) Programme Grants for Applied Research programme (project number RPPG-0606-1184). Journals Library.. Word count: xxxx. The final report from this programme will be published in the NIHR.
(5) Abstract The Sheffield Type 1 Diabetes Policy Model is a patient-level simulation model of type 1 diabetes and its associated complications, which was developed as part of the National Institute for Health Research Dose Adjustment for Normal Eating (DAFNE) research programme.. The aim of this paper is to describe the conceptual modelling, model. implementation, and model validation phases of the Sheffield Type 1 Diabetes Model development process. The model is highly flexible and has broad potential application to evaluate DAFNE, other diabetes structured education programmes, and other interventions for type 1 diabetes..
(6) Introduction Type 1 diabetes is a metabolic disorder characterised by an almost total deficiency in insulin that leads to higher than normal levels of glucose in patients’ blood (termed poor glycaemic control).. Once patients are diagnosed with type 1 diabetes they must remain on insulin. replacement therapy for their lifetime.. Type 1 diabetes is associated with long-term. microvascular complications (neuropathy, nephropathy, retinopathy, and macular oedema) and macrovascular complications (myocardial infarction (MI), heart failure (HF), stroke, and angina) which can lead to serious consequences such as limb amputation, blindness, disability and death. These diabetes-related complications account for most of the increased morbidity and mortality associated with type 1 diabetes1. The risk of long-term complications is related to patients’ glycaemic control, which is most commonly assessed using glycosylated haemoglobin (HbA1c), an average measure of blood glucose levels over time. Patients with type 1 diabetes are also at risk of acute complications: hypoglycaemia (excessively low blood glucose caused by taking too much insulin) and diabetic ketoacidosis (DKA) (high levels of ketones in the blood caused by high blood glucose levels). Both longterm and acute diabetic complications are associated with substantial healthcare costs and affect patients’ quality of life (QoL) and their mortality risk. In the UK it is recommended that all patients with type 1 diabetes are offered a structured education programme (SEP) to support their diabetes self-management2. The only SEP in the UK currently meeting the nationally agreed criteria is the Dose Adjustment For Normal Eating (DAFNE) course3. DAFNE is a five-day outpatient SEP aimed at providing adults with type 1 diabetes with the skills and confidence to estimate the carbohydrate content of meals and adjust their insulin doses to match food portions4. A randomised controlled trial (RCT) of 169 patients with type 1 diabetes demonstrated that DAFNE significantly improved HbA1c, dietary freedom and overall QoL compared with no DAFNE, without increasing the rate of hypoglycaemia5. A published cost-effectiveness analysis of DAFNE compared with no DAFNE suggested that the intervention was cost-effective and would pay for itself within five years6. The National Institute for Health Research (NIHR) funded a five-year research programme to investigate in more detail the factors affecting the success of DAFNE7. The programme, entitled “Improving Management of Type 1 Diabetes in the UK: The DAFNE Programme as a Research Test-bed”, was underpinned by health economic analyses. The health economic analyses underpinning this research programme included the development of a new health economic model to evaluate the cost-effectiveness of evolving forms of the DAFNE intervention. Sheffield.. This research was undertaken at the University of. The aim of this paper is to describe the conceptual modelling, model.
(7) implementation, and model validation phases of the Sheffield Type 1 Diabetes Model development process. The paper first outlines how the model was conceptually designed, then describes how it was implemented in the simulation software Simul8® and the key features of the model and its inputs. The results of the internal validation are provided. Finally the paper presents a discussion of the strengths and weaknesses of the Sheffield Type 1 Diabetes Model. Conceptual Modelling The conceptual modelling phase of the model development process included two workshops with clinical and social science experts in diabetes, a systematic review of published models of type 1 diabetes, and structured decision making by researchers at the University of Sheffield. An initial workshop (Workshop 1) was held in June 2009 with invited clinical diabetes specialists (including a nurse specialist) and the University of Sheffield DAFNE health economics team to understand the natural history of type 1 diabetes. The next stage of the conceptual modelling process was a systematic review of previously published costeffectiveness models of type 1 diabetes. A total of 65 papers, relating to 32 individual costeffectiveness models, were selected for inclusion in the review (details available from the authors on request).. A draft model structure including all potential diabetes-related. complications was then developed based on the systematic review of previous costeffectiveness models. In July 2010, the University of Sheffield DAFNE health economics team conducted a second workshop (Workshop 2) with clinical experts to discuss the results of the review and the proposed conceptual model. The final conceptual model was developed after discussions in the workshop and is as shown in Figure 1. Model Description The Sheffield Type 1 Diabetes Model is a flexible and comprehensive long-term simulated patient-level Markov model incorporating the most up-to-date methodologies (such as capturing parameter uncertainty, time profile of patient characteristics and including patient behaviour) to allow a number of cost-effectiveness evaluations. The Sheffield Type 1 Diabetes model consists of a series of sub-models simulating the progression of each of the diabetic complications, acute complications and mortality in a given population with type 1 diabetes. The model allows each simulated patient to develop multiple complications and for the incidence of these complications to be dependent upon simulated patients’ individual characteristics. The individual patient characteristics include demographics (age, gender and duration of diabetes), clinical variables (HbA1c, high density lipoprotein (HDL), smoking status, blood pressure and cholesterol), existing diabetes-related.
(8) complications and treatment status. The complications included in the model are nephropathy, retinopathy, neuropathy, severe hypoglycaemia, MI, stroke, HF and angina while the adverse events included hypoglycaemia and DKA, as shown in Figure 1. The progression of long-term diabetic complications are modelled using transition probabilities with an annual time cycle and the adverse events are modelled as annual incidence, for each individual patient based on their characteristics (patient behaviour can also be incorporated in the model by updating HbA1c and other variables over time). Each health state is associated with an annual cost and a utility value which is combined with the number of annual time cycles the patient spends in that health state to estimate costs and qualityadjusted life years (QALYs). Some disease progression events are associated with a one-off transition cost that is incurred in the transition year. Costs and QALYs are summed across time and patients to provide total and average cost and QALY estimates for use in costeffectiveness analyses. Figure 1. Structure of the Sheffield Type 1 Diabetes Model. Microvascular Complications The risk of development and progression of nephropathy, neuropathy and retinopathy are modelled according to event rates reported in published randomised controlled trials (RCTs) and observational studies.. Cohort Markov models were used to estimate annual.
(9) probabilities of transitioning between states within a particular complication, by combining data from multiple sources, assuming a reference HbA1c of 10%. The process was the same for all the microvascular complications (retinopathy, nephropathy, neuropathy and macular oedema) and full details of these methods are available from the authors on request.. For each microvascular complication, patients progress to the more severe health states within each annual time cycle according to the probabilities reported in Table 1. As the probabilities are estimated at the reference HbA1c of 10%, Eastman’s method8 was used to adjust the risk of background retinopathy, macular oedema, proliferative retinopathy, microalbuminuria, macroalbuminuria, and neuropathy for patients with different HbA1c levels (PHbA1c) using the formula: PHbA1c = PHbA1c=10(HbA1c/10)^. (Equation 1). Where is the baseline probabilities PHbA1c=10 are as shown in the Table 1 and the coefficients are as shown in the footnote of Table 1. The rest of the transition probabilities are assumed to be independent of HbA1c levels. Table 1. Annual probability of microvascular events Neuropathy Parameter. Base case value Annual transition probabilities for microvascular complications 0.0354 Healthy to clinically confirmed neuropathya Healthy to PAD with amputation 0.0003 Clinically confirmed neuropathy to PAD with 0.0154 amputation a. Base case value Annual transition probabilities for microvascular complications 0.0436 Healthy to microalbuminuriaa b 0.0037 Healthy to macroalbuminuria Healthy to ESRD 0.0002 Healthy to death from ESRD 3.3e-06 0.1565 Microalbuminuria to macroalbuminuriab Microalbuminuria to ESRD 0.0133 Microalbuminuria to death from ESRD 0.0004 Macroalbuminuria to ESRD 0.1579 Macroalbuminuria to death from ESRD 0.0070 ESRD to death from ESRD 0.0884 b. DCCT,9 Moss et al10 (WESDR). coefficient for neuropathy = 5.30. Nephropathy Parameter. a. Source(s). coefficient for microalbuminuria = 3.25 coefficient for macroalbuminuria = 7.95. Source(s). DCCT11, Wong el al12 (WESDR), UKPDS 3313.
(10) Retinopathy and macular oedema Parameter. Base case value Annual transition probabilities for microvascular complications 0.0454 Healthy to background retinopathya b 0.0013 Healthy to proliferative retinopathy 0.0012 Healthy to macular oedemac Healthy to blindness 1.9e-06 0.0595 Background retinopathy to proliferative retinopathyb c 0.0512 Background retinopathy to macular oedema Background retinopathy to blindness 0.0001 Proliferative retinopathy to blindness 0.0038 Macular oedema to blindness 0.0016 a b c. Source(s). WESDR XXII14. coefficient for background retinopathy = 10.10 coefficient for proliferative retinopathy = 6.30 coefficient for macular oedema = 1.20. Macrovascular Complications The risks of fatal and non-fatal macrovascular complications (MI, stroke, HF and angina) are modelled in three stages. First, the annual probability of experiencing any cardiovascular event, P_CVD, is estimated based on patients’ characteristics as per Cederholm et al’s 5year cardiovascular risk model15: P_CVD = 1-exp(-(-(ln(1 - 5year_CVD_risk))/5)*1). (Equation 2). Where 5year_CVD_risk is given by the equation 5year_CVD_risk = (1–0.97136exp [0.08426 × (duration–28.014) + 0.04742 × (age–duration– 16.601) + 0.80050 × (log(TC:HDL)–1.1470) + 1.27275 × (log(HbA1c(DCCT))–2.0605) + 1.20050 × (log(systolic BP)– 4.8598) + 0.56688 × (smoker–0.1483) + 0.41995 × (macroalbuminuria–0.1237) + 1.25506 × (previous CVD–0.0612)]. ). (Equation 3) The P_CVD probability is compared with a random number and if the random number is lower than the estimated probability then the patient is deemed to experience a cardiovascular event. Secondly, for those patients that experience an event, another random number is then used to determine what type of event it was (MI, stroke, HF or angina) using methods outlined in Palmer’s 2012 thesis16, based on data from the Diabetes Control and Complications Trial/Epidemiology of Diabetes Interventions and Complications (DCCT/EDIC) study17. Given a cardiovascular event, there is a 53% chance that is MI, 28% chance that it is angina, 12% chance that it is HF and 7% chance of stroke, as shown in Table 2. Thirdly, if the event experienced is an MI, stroke or HF, further random numbers are then used to.
(11) determine whether the event is fatal using methods outline in Palmer’s 2012 thesis16 and as shown in Table 3. Table 2 Probability of different cardiovascular events Parameter MI. Base case value 0.53. Gamma Distribution alpha beta 1 0.0053. Stroke. 0.07. 1. 0.0007. Angina. 0.28. 1. HF. 0.12. 1. 0.0028 0.00126. Source(s). DCCT/EDIC17. Table 3. Probability of dying from cardiovascular events Parameter. Base case value. Gamma Distribution alpha beta 1 0.00393. Source(s). MI death in hospital: Men. 0.3930. MI death in hospital: Women. 0.3640. 1. 0.00364. Sonke et al18. MI death within one year: Aged < 0.1522. 1. 0.00152. Malmberg et al19. 1. 0.00186. Malmberg et al19. 1. 0.00250. Malmberg et al19. Sonke et al18. 65 years MI death within one year: Aged 65- 0.1860 75 years MI death within one year: Aged > 0.2508 75 years Stroke death within 30 days. 0.1240. 1. 0.00124. Eriksson et al20. Stroke death within one year. 0.1063. 1. 0.00106. DCCT/EDIC17. HF death within one year. 0.0570. 1. 0.00057. Anselmino et al21. Acute Complications Two acute complications are simulated in the Sheffield Type 1 Diabetes Model: severe hypoglycaemia (defined as a hypoglycaemic event that the person with type 1 diabetes is unable to treat themselves) and DKA. The model parameters on the incidence of these two events were estimated from the DAFNE Research Database and the original DAFNE vs. no DAFNE RCT dataset22. Negative binomial models were developed to predict the annual rates and the results of the models are presented in Table 4. These models were inputted into R software to generate 10,000 samples of the number of severe hypoglycaemic and DKA episodes for patients with HbA1c values from 4% to 16% in 0.1% increments.. The. simulated samples were used to define probability distributions which random numbers were compared to within the model in order to determine how many events each simulated patient.
(12) had in each year (based on their HbA1c value and whether they had received DAFNE or not). Full details of these methods are available from the authors on request. Table 4: Negative binomial models of the annual number of severe hypoglycaemic and DKA episodes Coefficient. Standard error. 95% confidence interval. 0.928. 0.553. (-0.155, 2.012). -0.113. 0.064. (-0.259, -0.006). -8.108. 1.097. (-10.259, -5.958). 0.617. 0.115. (0.392, 0.842). Severe hypoglycaemia Intercept HbA1c. 1H 2H. DKA Intercept HbA1c a. 1K 2K. the negative binomial model is Log(number of events) = Intercept B 1 + ( 2*HbA1c) + error. Mortality Patients can also die due to other causes (than due to ESRD and CVD) and this other cause mortality is modelled based on UK Interim Life Tables from 2008-1023. The model compares random numbers to gender- and age-specific annual probabilities of death and if the random number is lower than the probability of death then the patient is simulated to be dead. The model allows for other life tables to be selected e.g. there is an option to select US mortality data used in the CORE model24 or mortality rates from the EAGLE model25. Treatment Effectiveness HbA1c is the primary method of accounting for treatment effects in the model. However, intervention effects on other risk factors such as blood pressure, cholesterol or severe hypoglycaemia can also be incorporated and the profiles of the risk factors over time can be updated annually. HbA1c change as a result of an intervention has an impact on the risk of developing several microvascular complications and this effect is modelled based on Eastman’s method8 of adjusting the risk for changes in HbA1c levels as outlined above. For macrovascular complications, the coefficients for HbA1c, HDL, smoking status, blood pressure or cholesterol used in Cederholm et al15 were used to adjust the probability of any cardiovascular event.. Finally, the effect of interventions on outcomes such as. hypoglycaemia and DKA can be input directly into the model by the user. Utilities The model calculates long-term QALYs by using utility values for the health states from the literature, reported in Table 5. Each health state is associated with a disutility value.
(13) (negative) which is added to the baseline utility to estimate the utility in the given health state.. In case of multiple complications, the utilities are estimated by aggregating the. disutilities of the multiple complications to the baseline utility. The lifetime QALYs for each patient are estimated based on patients’ life expectancy and their corresponding annual utilities. The model has the flexibility to use alternative utility values as inputted by the model user. Table 5:. Base case utility parameters. Health state or event. Utility. Beta distribution Alpha. Source(s). Beta. Baseline utility values Male with type 1 diabetes and no 0.672. 3022.176 1475.11. Coffey26. Gamma Distribution. Source(s). complications Utility decrements Complications or covariates. Disutility. Alpha Female with type 1 diabetes and no -0.033. Beta. 17.01563 0.001939. Coffey26. complications Blindness. -0.208. 256. 0.000813. Assumption. Macroalbuminuria. -0.017. 2.89. 0.005882. Coffey26. ESRD. -0.023. 0.725652 0.031696. Coffey26. Clinically confirmed neuropathy. -0.055. 30.25. 0.001818. Coffey26. PAD with amputation. -0.116. 25.43667 0.004561. Coffey26. Background retinopathy. -. -. -. Assumption. Proliferative retinopathy. -. -. -. Assumption. Macular oedema. -. -. -. Assumption. MI (assumed equal to HF). -0.058. 6.950413 0.008345. Coffey26. Stroke. -0.018. 0.669421 0.026889. Coffey26. HF. -0.058. 6.950413 0.008345. Coffey26. Angina. -0.090. 24.00912 0.003749. UKPDS 62. Severe hypoglycaemia. -0.071. Samples. Walters. Samples. et. 27. al DKA (assumed equal to severe hypo but without ongoing utility decrement due to fear of hypos). -0.001. Samples. Samples. Walters al27. et.
(14) Costs The model calculates long-term costs by using health state costs values from the literature, as presented in Table 6.. Each health state is associated with an annual cost which is. combined with the number of annual time cycles the patient spends in that health state to estimate the costs. In case of multiple complications, the costs are estimated by aggregating the annual costs of the different complications. Some disease progression events are also associated with a one-off transition cost that is incurred in the transition year. All costs have been inflated to 2010/11 prices using Personal Social Services Research Unit inflation indices28. The model has the flexibility to use alternative cost profiles as inputted by the model user. Table 6:. Base case health state and transition costs Mean Costs. Gamma Distribution. Source. Alpha 100 100 100. Beta 0.34 0.34 232.75. £258 £258. 100 100. 2.58 2.58. Clinical Neuropathy Diab foot syndrome. £2,713. 100. 27.13. Currie et al32 Assumed equal to clinical confirmed neuropathy NHS Reference Costs31. PAD with amputation (year 1). £6,878. 100. 68.78. NHS Reference Costs31. £418 £138 £630 £630. 100 100 100 100. 4.18 1.38 6.30 6.30. £1,509 £494 £6,465 £6,465 £6,465 £861 £2,001 £4,154 £4,154 £532 £5,414 £3,637 £1,117. 100 100 100 100 100 100 100 100 100 100 100 100 100. 15.09 4.94 64.65 64.65 64.65 8.61 20.01 41.54 41.54 5.32 54.14 36.37 11.17. Microalbuminuria (ongoing) Macroalbuminuria (ongoing) ESRD (ongoing) Clin Conf Neuropathy. PAD with amputation (ongoing) Background Retinopathy Proliferative Retinopathy Macular edema Blindness (year 1) Blindness (ongoing) First MI (year 1) Second MI Final MI MI (ongoing) Fatal MI First Stroke (year 1) Second Stroke First Stroke (ongoing) Fatal Stroke HF (year 1) HF (ongoing). £34 £34 £23,275. BNF29, McEwan et al30 BNF29, McEwan et al30 NHS Reference Costs31. McEwan et al30 McEwan et al30 McEwan et al30 Assumed equal to proliferative retinopathy UKPDS 6533 UKPDS 6533 UKPDS 6533 UKPDS 6533 UKPDS 6533 UKPDS 6533 UKPDS 6533 UKPDS 6533 UKPDS 6533 UKPDS 6533 UKPDS 6533 UKPDS 6533 UKPDS 6533.
(15) Fatal HF Angina (year 1) Angina (ongoing) Hypos Hypos with Comma. Hypos with Hospitalisation. £3,637 £3,236 £906 £178 £702. 100 100 100 100 100. 36.37 32.36 9.06 1.78 7.02. £702. 100. 7.02. UKPDS 6533 UKPDS 6533 UKPDS 6533 Our calculation Assumed equal to hypo w/ hosp NHS Reference Costs31 NHS Reference Costs31. DKA with Hospitalisation Cost of a diabetic patient with no complications. £1,333. 100. 13.33. £4,212. 100. 42.12. UKPDS 6533. Other Model details The model was developed in line with the modelling good practice guidelines34, recommendations from the American Diabetes Association35 and published checklists for economic evaluation36,37. The model uses an annual discount rate of 3.5% as default (for both costs and QALYs, as recommended by NICE38). The model takes a health service perspective and uses a lifetime horizon (i.e. until all simulated patients have died) as default but the perspective and time horizon are flexible and can be set by the model user. The model is capable of performing probabilistic sensitivity analysis (PSA) allowing the effects of parameter uncertainty to be captured and the likelihood that interventions are cost-effective to be reported. The decision uncertainty is estimated using probability distributions (or a collection of random samples) for uncertain parameters. Where parameters were correlated and the covariance matrix was known, the random samples were drawn from a multivariate distribution. Model Flexibility The model, programmed in Simul8® software, was developed in a flexible manner that allows alternative sets of input data. The user can select whether to perform a deterministic analysis or conduct PSA, whereby model parameters are sampled from probability distributions. The model also has several option dialogs that allow the user to change the time horizon, discount rates for costs and QALYs, patient cohort characteristics, cohort size, treatment effects, and cost and utility sources. The Sheffield Type 1 Diabetes Model is highly flexible to allow for a large number of differing cost-effectiveness analyses to be undertaken. Model Outputs The model also allows tracking the history of each of the patients every year which allows easy verification and validation of the model. This includes the patient characteristics (i.e. HbA1c, SBP, HDL, etc), incidence of acute complications (i.e. hypos and DKA), and.
(16) microvascular and macrovascular complication status (i.e. disease progression) for each year the patient is alive. The aggregated numbers of patients in different health states are output each year and the total numbers of each event are also output at the end of the lifetime horizon. The costs and utility values, including the split of costs and disutilities by complication, are output for each patient for every year they are alive. The total discounted costs and QALYs are also output at the end of the lifetime horizon. When performing PSA, for the sake of efficiency, the model does not track the history of each patient every year but outputs the total costs, QALYs and the numbers of events in each complication for each PSA run. Model Verification Internal verification of the model code (visual logic in Simul8®) was conducted throughout the model implementation process. Patient characteristics and complication statuses were checked to ensure that they were changing as expected, and that patients were following expected routes. The costs and utility value outputs each year were checked against the patient status outputs for face validity. The aggregated outputs were also cross checked against the sum of individual patient outputs. Second-order validation was also conducted, whereby the risk model was internally validated against the data from which it was estimated. Results The results of second-order validation, which compared the model results with the data from the studies used to build the model, are as shown in Table 7. For microvascular complications, the normalised differences between model results and the published data ranged between 0-15%, except for the deaths from ESRD (which is more than 50%, but can be attributed to low event rates) and neuropathy events (~ 25%), with most difference less than 10%. For macrovascular complications, the normalised differences between model results and the published data ranged between 0-10%, with most differences less than 5%.. Table 7. Results of second order validation Microvascular Complication Nephropathy. Source. Observed incidence (%). Modelled incidence (%). Microalbuminuria. DCCT11. 20%. 17%. Macroalbuminuria. Wong el al12 (WESDR). 33%. 27%.
(17) ESRD. Wong el al12 (WESDR). 20%. 18%. Death from ESRD Retinopathy. UKPDS 3313. 0.26%. 0.11%. BDR. WESDR XXII14. 80%. 64%. PDR. WESDR XXII14. 39%. 40%. ME. WESDR XXII14. 26%. 18%. Blindness. WESDR XXII14. 2.3%. 2.3%. Neuropathy. DCCT,9. 9.3%. 11.9%. Amputation. Moss et al10 (WESDR). 9.6%. 9.5%. Macrovascular Complication. Source. MI. Cederholm et al,15 Palmer’s thesis16. Neuropathy. Observed % of total events. Modelled % of total events. 53%. 52%. 7%. 7%. 12%. 13%. 28%. 29%. 5.41%. 5.61%. 15. Angina. Cederholm et al, Palmer’s thesis16 Cederholm et al,15 Palmer’s thesis16 Cederholm et al,15 Palmer’s thesis16. All CVD. Cederholm et al15. Stroke HF. Discussion and Conclusions The Sheffield Type 1 Diabetes Model has several key strengths. Firstly, the model is based on a structured conceptual modelling process that included input from multidisciplinary experts in the fields of clinical diabetes, psychology, diabetes education, and simulation modelling.. This structured process ensured that the development of the model was. evidence-based and that the model is fit for purpose from a number of disciplinary perspectives.. Secondly, the model is highly flexible, allowing users to specify the. characteristics of simulated patients, the time horizon, the cohort size, how treatment effects are accounted for, what outcomes are tracked by the model, and whether to run the model deterministically or probabilistically.. This high level of flexibility allows the model to be. adapted to the user’s particular research question, setting, or population of interest and broadens the model’s potential applications. Thirdly, the model is a patient-level simulation.
(18) which offers the advantage of being able to account for individual differences between patients. Fourthly,. the model allows for. patients’ psychological and behavioural. characteristics and their impact on treatment effectiveness to be incorporated into analyses. These two features of the model are particularly useful for investigating heterogeneous populations or subgroups. Finally, the model is structured to facilitate probabilistic sensitivity analysis which accounts for uncertainty in the model parameters and is recommended by several health technology assessment agencies including NICE38. Despite its many advantages the Sheffield Type 1 Diabetes Model also has some limitations. The Sheffield Type 1 Diabetes Policy Model used published data from non-UK settings to define risk of long-term complications, some of which are now very old. The risk of long-term macrovascular complications is dependent mainly on HbA1c and the effect of other risk factors is not captured, which might cause bias when evaluating interventions that affect risk factors other than HbA1c. Although the uncertainty in most of the parameters is incorporated into the model, uncertainty in some parameters (e.g. coefficients of the risk equations) is not captured. Future Research/Planned Analysis The Sheffield Type 1 Diabetes model was developed as part of the NIHR DAFNE research programme and several model-based evaluations are planned as part of that programme. Firstly, the model will be used to update the cost-effectiveness results reported by Shearer et al6 to include the effects of DAFNE on long-term incidence of macrovascular as well as microvascular complications.. Secondly, the model will be used to evaluate DAFNE. delivered one day per week over five weeks compared with original DAFNE (five consecutive days) and thirdly, to evaluate DAFNE plus insulin pumps versus DAFNE plus MDI.. The Sheffield Type 1 Diabetes model can also be used to evaluate any (i.e. non-. DAFNE) interventions for type 1 diabetes. There are also plans to re-estimate the risk equations from longitudinal data from DAFNE research database and the long-term followup data from DCCT/EDIC. Several additions and adaptations to the model are also planned. Planned changes include addition of alternative cost and utility input databases from DAFNE research database and/or RCTs. Summary In summary, the Sheffield Type 1 Diabetes Model offers a new whole disease model of type 1 diabetes and its associated complications.. The model development process was. evidence-based and in consultation with multi-disciplinary experts.. The model is highly. flexible and has broad potential application to evaluate DAFNE, other diabetes structured.
(19) education programmes, and other interventions for type 1 diabetes.. The model is under. constant development and updating and several adaptations are planned. Acknowledgments: This article discusses independent research funded by the National Institute for Health Research (NIHR) under its Programme Grants for Applied Research scheme (RP-PG-0606-1184). The views expressed in this presentation are those of the author(s) and not necessarily those of the NHS, the NIHR or the Department of Health. References 1. Daneman, D. Type 1 diabetes. 2006; 367:847â€“58. 2. National Institute for Health and Clinical Excellence. Structured education. 2011. 3. Department of Health, Diabetes, U.K. Structured patient education in diabetes: Report from the Patient Education Working Group. 2005. 4. DAFNE. What is DAFNE? 2011; 2011. 5. DAFNE Study Group. Training in flexible, intensive insulin management to enable dietary freedom in people with type 1 diabetes: dose adjustment for normal eating (DAFNE) randomised controlled trial. 2002; 325:746. 6. Shearer, A., Bagust, A., Sanderson, D., Heller, S., Roberts, S. Cost-effectiveness of flexible intensive insulin management to enable dietary freedom in people with Type 1 diabetes in the UK. 2004; 21:460-467. 7. DAFNE. Fact Sheet Eleven. 2011; 2011:U07. 8. Eastman, R.C., Javitt, J.C., Herman, W.H., Dasbach, E.J., Zbrozek, A.S., Dong, F. et al. Model of complications of NIDDM. I. Model construction and assumptions. Diabetes Care 1997; 20(5):725-734. 9.. The effect of intensive diabetes therapy on the development and progression of neuropathy. The Diabetes Control and Complications Trial Research Group. Ann Intern Med 1995; 122(8):561-568.. 10.. Moss, S.E., Klein, R., Klein, B.E., Wong, T.Y. Retinal vascular changes and 20-year incidence of lower extremity amputations in a cohort with diabetes. Arch Intern Med 2003; 163(20):2505-2510.. 11. Effect of intensive therapy on the development and progression of diabetic nephropathy in the Diabetes Control and Complications Trial. The Diabetes Control and Complications (DCCT) Research Group. Kidney Int 1995; 47(6):1703-1720. 12. Wong, T.Y., Shankar, A., Klein, R., Klein, B.E. Retinal vessel diameters and the incidence of gross proteinuria and renal insufficiency in people with type 1 diabetes. Diabetes 2004; 53(1):179-184..
(20) 13. Intensive blood-glucose control with sulphonylureas or insulin compared with conventional treatment and risk of complications in patients with type 2 diabetes (UKPDS 33). UK Prospective Diabetes Study (UKPDS) Group. Lancet 1998; 352(9131):837-853. 14. Klein, R., Knudtson, M.D., Lee, K.E., Gangnon, R., Klein, B.E. The Wisconsin Epidemiologic Study of Diabetic Retinopathy: XXII the twenty-five-year progression of retinopathy in persons with type 1 diabetes. Ophthalmology 2008; 115(11):1859-1868. 15. Cederholm, J., Eeg-Olofsson, K., Eliasson, B., Zethelius, B., Nilsson, P.M., Gudbjornsdottir, S. Risk prediction of cardiovascular disease in type 2 diabetes: a risk equation from the Swedish National Diabetes Register. Diabetes Care 2008; 31(10):2038-2043. 16. Palmer J. Cost-effectiveness of angiotensin-converting enzyme inhibitors for the prevention of cardiovascular and end-stage renal disease in patients with type 1 diabetes [ 2012. 17. Nathan, D.M., Cleary, P.A., Backlund, J.Y., Genuth, S.M., Lachin, J.M., Orchard, T.J. et al. Intensive diabetes treatment and cardiovascular disease in patients with type 1 diabetes. N Engl J Med 2005; 353(25):2643-2653. 18. Sonke, G.S., Beaglehole, R., Stewart, A.W., Jackson, R., Stewart, F.M. Sex differences in case fatality before and after admission to hospital after acute cardiac events: analysis of community based coronary heart disease register. BMJ 1996; 313(7061):853-855. 19. Malmberg, K., Ryden, L., Efendic, S., Herlitz, J., Nicol, P., Waldenstrom, A. et al. Randomized trial of insulin-glucose infusion followed by subcutaneous insulin treatment in diabetic patients with acute myocardial infarction (DIGAMI study): effects on mortality at 1 year. J Am Coll Cardiol 1995; 26(1):57-65. 20.. Eriksson, S.E., Olsson, J.E. Survival and recurrent strokes in patients with different subtypes of stroke: a fourteen-year follow-up study. Cerebrovasc Dis 2001; 12(3):171180.. 21.. Anselmino, M., Malmberg, K., Ohrvik, J., Ryden, L. Evidence-based medication and revascularization: powerful tools in the management of patients with diabetes and coronary artery disease: a report from the Euro Heart Survey on diabetes and the heart. Eur J Cardiovasc Prev Rehabil 2008; 15(2):216-223.. 22. DAFNE NIHR Research Group. Personal communication: Unpublished data. 2006. 23. UK Interim Life Tables 2008-10. 2010; available from http://www.ons.gov.uk/ons/rel/lifetables/interim-life-tables/2008-2010/sum-ilt2008-10.html 24. Palmer, A.J., Roze, S., Valentine, W.J., Minshall, M.E., Foos, V., Lurati, F.M. et al. The CORE Diabetes Model: Projecting long-term clinical outcomes, costs and costeffectiveness of interventions in diabetes mellitus (types 1 and 2) to support clinical and reimbursement decision-making. 2004; 20 Suppl 1:S5-26..
(21) 25. Mueller, E., Maxion-Bergemann, S., Gultyaev, D., Walzer, S., Freemantle, N., Mathieu, C. et al. Development and validation of the Economic Assessment of Glycemic Control and Long-Term Effects of diabetes (EAGLE) model. 2006; 8:219-236. 26. Coffey, J.T., Brandle, M., Zhou, H., Marriott, D., Burke, R., Tabaei, B.P. et al. Valuing health-related quality of life in diabetes. 2002; 25:2238-2243. 27.. Walters, N., Gordois, A., Brown, A., Lindsay, P., Gonzalo, F., Comas, S. Quantifying the impact of fear of hypoglycemia on quality of life. 2006; available from http://www.journals.elsevierhealth.com/periodicals/jval/article/S10983015%2810%2963316-5/abstract. 28. Personal Social Services. unit costs of health and social care 2010. 2010. 29. Joint Formulary Committee. British National Formulary 61. British Medical Association and Royal Pharmaceutical Society, London; 2011. 30.. McEwan, P., Poole, C.D., Tetlow, T., Holmes, P., Currie, C.J. Evaluation of the costeffectiveness of insulin glargine versus NPH insulin for the treatment of type 1 diabetes in the UK. 2007; 23:2007.. 31. Department of Health. NHS reference costs 2007-08. 2009. London, UK; 2009. 2008; 32.. Currie, C.J., Poole, C.D., Woehl, A., Morgan, C.L., Cawley, S., Rousculp, M.D. et al. The financial costs of healthcare treatment for people with Type 1 or Type 2 diabetes in the UK with particular reference to differing severity of peripheral neuropathy. Diabet Med 2007; 24(2):187-194.. 33.. Clarke, P., Gray, A., Legood, R., Briggs, A., Holman, R. The impact of diabetes-related complications on healthcare costs: results from the United Kingdom Prospective Diabetes Study (UKPDS Study No. 65). Diabet Med 2003; 20(6):442-450.. 34.. Caro, J.J., Briggs, A.H., Siebert, U., Kuntz, K.M. Modeling good research practices-overview: a report of the ISPOR-SMDM Modeling Good Research Practices Task Force-1. Med Decis Making 2012; 32(5):667-677.. 35. Guidelines for computer modeling of diabetes and its complications. Diabetes Care 2004; 27(9):2262-2265. 36. Eddy, D.M. Technology assessment: the role of mathematical modelling. In: Institute of Medicine, eds. Assessing Medical Technologies. National Academy Press; Washington DC: 1985. 37. Drummond, M.F., Jefferson, T.O. Guidelines for authors and peer reviewers of economic submissions to the BMJ. 1996; 313:275-283. 38. National Institute for Health and Clinical Excellence. Guide to the methods of technology appraisal. 2008..